Improving Photometric Redshifts by Merging Probability Density Functions from Template-Based and Machine Learning Algorithms

Abstract

This study aims to improve the photometric redshifts (photo-$z$s) of galaxies by integrating two contemporary methods: template-fitting and machine learning. Finding the synergy between these two methods was not a high priority in the past, but now that our computer processing power and observational accuracy have increased, we deem it worth investigating. We compared two methods to improve galaxy photometric redshift estimations by using the algorithms ANNz2 and BPz on different photometric and spectroscopic samples from the Sloan Digital Sky Survey (SDSS). We find that the photometric redshift performance of ANNz2 (machine learning) is better than that of BPz (galactic templates), and with the utilisation of the merging technique we introduced, we see that there is an improvement in photo-$z$ when the two strategies are consolidated, providing improvements in ${\sigma }_{RMS}$ and ${\sigma }_{68}$ up to [0.0265, 0.0222] in the LRG sample and [0.0471, 0.0471] in the Stripe-82 Sample. This simple demonstration can be used for photo-$z$s of galaxies in fainter and deeper sky surveys, and future work is required to prove its viability in these samples.

14 pages, 5 figures

References in corpus (5)

• Photometric Redshifts for Next-Generation Surveys
• Photometric redshifts for the Kilo-Degree Survey. Machine-learning analysis with artificial neural networks
• On the realistic validation of photometric redshifts, or why Teddy will never be Happy
• Morpho-z: improving photometric redshifts with galaxy morphology
• Photometric redshifts with machine learning, lights and shadows on a complex data science use case